U.S. patent number 6,735,248 [Application Number 10/071,451] was granted by the patent office on 2004-05-11 for fast converging equalizer for a demodulator.
This patent grant is currently assigned to International Business Machines Corporation. Invention is credited to Jian Gu, Jianping Pan.
United States Patent |
6,735,248 |
Gu , et al. |
May 11, 2004 |
Fast converging equalizer for a demodulator
Abstract
A method and a system for configuring and initializing an
equalizer. The equalizer equalizes a signal received over a
propagation channel. The received signal includes symbols, a
synchronizing signal and is formed by one or more rays including a
dominant ray. The method comprising the following: (a) generating a
multi-path intensity profile of the propagation channel from the
received signal; (b) determining an arrival time of the dominant
ray and a propagation channel response corresponding to the arrival
time based on the multi-path intensity profile; (c) decimating the
received signal synchronously with the arrival time of the dominant
ray to provide signal samples to the equalizer; (d) configuring the
equalizer based on the multi-path intensity profile to deactivate
some of the coefficients of the equalizer; and (e) initializing
active coefficients of the equalizer based on the propagation
channel response.
Inventors: |
Gu; Jian (San Diego, CA),
Pan; Jianping (San Diego, CA) |
Assignee: |
International Business Machines
Corporation (Armonk, NY)
|
Family
ID: |
23013567 |
Appl.
No.: |
10/071,451 |
Filed: |
February 8, 2002 |
Related U.S. Patent Documents
|
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
Issue Date |
|
|
266193 |
Mar 10, 1999 |
|
|
|
|
Current U.S.
Class: |
375/232;
375/350 |
Current CPC
Class: |
H04L
25/03038 (20130101); H04L 2025/03375 (20130101); H04L
2025/03477 (20130101) |
Current International
Class: |
H04L
25/03 (20060101); H03H 007/30 () |
Field of
Search: |
;375/323,232,231,233,234,343,346,350 |
References Cited
[Referenced By]
U.S. Patent Documents
Other References
John G. Proakis, Digital Communications, Jan. 26, 1983, McGraw
Hill, United States. .
E. Eleftherious & D.Falconer, Adaptive Equalization Techniques
for HF Cahnnels, 2/87, IEEE Journal on Selected Areas in
Communications, vol. SAC-5, No. 2. .
R. D'Avella, L. Moreno, M. Sant'Agostino, an Adaptive MLSE Receiver
for TDMA Digital Mobile Radio, 1/89, IEEE Journal on Selected Areas
in Communications, vol. 7, No. 1. .
Modified Viterbi Equaliser for Mobile Radio Channels Having Large
Multipath Delays, Sep. 14, 1989, Electronics Letters, vol. 25, No.
19. .
J. Proakis, Adaptive Equalization for TDMA Digital Mobile Radio,
5/91, IEEE Journal on Selected Areas in Communications, vol. 40,
No. 2. .
G. D'Aria, R. Piermarini, V. Zingarelli, Fast Adative Equalizers
for Narrow-Band TDMA Mobile Radio, 5/91, IEEE Journal on Selected
Areas in Communications, vol. 40, No. 2. .
E. Delre, G. Benelli, G. Castellini, R. Fantacci, L. Pierucci, L.
Pogliani, Design of a Digital MLSE Receiver for Mobile Radio
Communications, 12/91, IEEE Global Telecommunications Conference,
Phoenix, Arizona. .
G. Benelli, A. Fioravanti, A. Garzelli, P. Matteini, Some Digital
Receivers for the GSM Pan-European Cellular Communication System,
6/94, IEEE Proc.-Commun., vol. 141, No. 3. .
G. Benelli, A Garzelli, F. Salvi, Simplified Viterbi Processors for
the GSM Pan-European Cellular Communication System, 11/94, IEEE
Transactions on Vehicular Technology, vol. 43, No. 4. .
European Telecommunications Standards Institute, Digital Cellular
Telecommunications System (Phase 2+); Radio Transmission and
Reception, 7/96, France..
|
Primary Examiner: Vo; Don N.
Attorney, Agent or Firm: Blakely, Sokoloff, Taylor &
Zafman LLP
Parent Case Text
CROSS-REFERENCE TO RELATED APPLICATIONS
This is a continuation application of U.S. patent application Ser.
No. 09/266,193 filed Mar. 10, 1999 now abandoned.
Claims
What is claimed is:
1. A method for configuring and initializing an equalizer, the
equalizer equalizing a signal received over a propagation channel,
the received signal having symbols, including a synchronizing
signal and being formed by one or more rays including a dominant
ray, the method comprising the operations of: (a) generating a
multi-path intensity profile of the propagation channel from the
received signal; (b) determining an arrival time of the dominant
ray and a propagation channel response corresponding to the arrival
time based on the multi-path intensity profile; (c) decimating the
received signal synchronously with the arrival time of the dominant
ray to provide signal samples to the equalizer, the equalizer
having coefficients; (d) configuring the equalizer based on the
multi-path intensity profile to deactivate some of the
coefficients; and (e) initializing active coefficients of the
equalizer based on the propagation channel response.
2. The method of claim 1 wherein the received signal is a digital
signal corresponding to a baseband signal.
3. The method of claim 1 further comprising the operation of
computing a nominal time, the nominal time being an average of
selected arrival times of dominant rays in a plurality of past
received signals.
4. The method of claim 3 further comprising the operation of
storing the received signal in a sample buffer at a rate greater
than one sample per symbol, a starting time of the storing of the
received signal being regulated by the nominal time.
5. The method of claim 3 further comprising the operation of
computing an average power of the synchronizing signal by computing
an average of squared magnitudes of samples of the synchronizing
signal.
6. The method of claim 5 further comprising the operation of
computing a signal-to-noise ratio based on the average power of the
synchronizing signal and a portion of the multi-path intensity
profile, the signal-to-noise ratio being used for selecting the
selected arrival times of dominant rays.
7. The method of claim 1 wherein, in operation (c), the signal
samples are provided to the equalizer at a rate of one signal
sample per symbol.
8. The method of claim 1 wherein operation (a) comprises the
operations of: correlating the received signal with a sequence
corresponding to the synchronizing signal to produce complex
correlator outputs, each of the complex correlator outputs
corresponding to a time index; and computing squared magnitudes of
the complex correlator outputs, each of the squared magnitudes
corresponding to a time index, the squared magnitudes forming the
multi-path intensity profile.
9. The method of claim 8 wherein operation (b) comprises the
operations of: determining a maximum value of the squared
magnitudes of the complex correlator outputs; and determining a
time index corresponding to the maximum value and the propagation
channel response, the time index being the arrival time of the
dominant ray, the propagation channel response being a complex
correlator output corresponding to the time index.
10. The method of claim 1 wherein operation (d) comprises the
operations of: (1) determining a number of coefficients; (2)
determining an equalization characteristic of the multi-path
intensity profile; and (3) deactivating some of the coefficients
when the equalization characteristic is less than a first
threshold.
11. The method of claim 1 wherein operation (e) comprises the
operations of: equating the active coefficients with zero, except
for a first active coefficient; and equating the first active
coefficient with the propagation channel response.
12. The method of claim 10 wherein the operation of deactivating
some of the coefficients when the equalization characteristic is
less than a predetermined threshold comprises the operations of:
(i) computing a set of test values from the multi-path intensity
profile, the test values corresponding one-to-one to the
coefficients, the coefficients being ordered; (ii) deactivating a
coefficient with a lowest order from the coefficients if a
corresponding test value is smaller than a second threshold; (iii)
repeating operation (ii) for remaining coefficients until a test
value corresponding to a coefficient with a lowest order is greater
than or equal to the second threshold; (iv) deactivating a
coefficient with a highest order from the coefficients if a
corresponding test value is smaller than a third threshold; and (v)
repeating operation (iv) for remaining coefficients until a test
value corresponding to a coefficient with a highest order is
greater than or equal to the third threshold.
13. The method of claim 12 wherein each of the test values is equal
to a partial sum of squared magnitudes of the complex correlator
outputs over a portion of the multi-path intensity profile, a
middle test value corresponding to a middle portion of the
multi-path intensity profile, the middle test value being one of
the test values.
14. The method of claim 13 wherein the equalization characteristic
is equal to a ratio of the sum of the test values to the middle
test value.
15. The method of claim 13 wherein the first threshold is a number
greater than 1.
16. The method of claim 13 wherein the second threshold is equal to
a product of the middle test value and a number smaller than 1.
17. The method of claim 13 wherein the third threshold is equal to
a product of the middle test value and a number smaller than 1.
18. The method of claim 1 further comprising the operation of
updating the active coefficients of the equalizer.
19. The method of claim 18 wherein the operation of updating the
active coefficients of the equalizer comprises the operation of
training the coefficients using a least mean squares algorithm.
20. An system for configuring and initializing an equalizer, the
equalizer equalizing a signal received over a propagation channel,
the received signal having symbols, including a synchronizing
signal and being formed by one or more rays including a dominant
ray, the system comprising: (a) a profile generator generating a
multi-path intensity profile of the propagation channel from the
received signal; (b) a determination module coupled to the profile
generator, the determination module determining an arrival time of
the dominant ray and a propagation channel response corresponding
to the arrival time based on the multi-path intensity profile; (c)
a decimator coupled to the determination module, the decimator
decimating the received signal synchronously with the arrival time
of the dominant ray to produce signal samples; (d) the equalizer,
having coefficients, coupled to the decimator to receive the signal
samples, the equalizer equalizing the signal samples to produce a
filtered signal; (e) a configuring module coupled to the equalizer,
the configuring module configuring the equalizer based on the
multi-path intensity profile; and (f) an initializing module
coupled to the equalizer, the initializing module initializing the
coefficients of the equalizer based on the propagation channel
response.
21. The system of claim 20 wherein the received signal is a digital
signal corresponding to a baseband signal.
22. The system of claim 20 further comprising a time recovery
module, the time recovery module computing a nominal time, the
nominal time being an average of selected arrival times of dominant
rays included in a plurality of past received signals.
23. The system of claim 22 further comprising a sample buffer
coupled to the time recovery module, the sample buffer storing the
received signal at a rate greater than one sample per symbol, a
starting time of the storing of the received signal being regulated
by the nominal time.
24. The system of claim 22 further comprising a power measurement
module, the power measurement module computing an average power of
the synchronizing signal by computing an average of squared
magnitudes of samples of the synchronizing signal.
25. The system of claim 24 further comprising a signal-to-noise
ratio estimator, the signal-to-noise ratio estimator computing a
signal-to-noise ratio based on the average power of the
synchronizing signal and a portion of the multi-path intensity
profile, the signal-to-noise ratio being used for selecting the
selected arrival times of dominant rays.
26. The system of claim 20 wherein the decimator produces the
signal samples at a rate of one signal sample per symbol.
27. The system of claim 20 wherein the profile generator comprises:
a correlator correlating the received signal with a sequence
corresponding to the synchronizing signal and generating complex
correlator outputs, each of the complex correlator outputs
corresponding to a time index; and a squared magnitude module
coupled to the correlator, the squared magnitude module computing
squared magnitudes of the complex correlator outputs, each of the
squared magnitudes corresponding to a time index, the squared
magnitudes forming the multi-path intensity profile.
28. The system of claim 27 wherein the determination module
comprises: a peak detector determining a maximum value of the
squared magnitudes of the complex correlator outputs; and a time
determinator determining a time index corresponding to the maximum
value and the propagation channel response, the time index being
the arrival time of the dominant ray, the propagation channel
response being a complex correlator output corresponding to the
time index.
29. The system of claim 27 wherein the configuring module
determines a number of coefficients and comprises: (1) a first
module determining an equalization characteristic of the multi-path
intensity profile; and (2) an optimizer deactivating some of the
coefficients when the equalization characteristic is less than a
first threshold.
30. The system of claim 27 wherein the initializing module equates
the active coefficients with zero, except for a first active
coefficient, and equates the first active coefficient with the
propagation channel response.
31. The system of claim 29 wherein the optimizer performs the
following: (i) computing a set of test values from the multi-path
intensity profile, the test values corresponding one-to-one to the
coefficients, the coefficient being ordered; (ii) deactivating a
coefficient with a lowest order from the coefficients if a
corresponding test value is smaller than a second threshold; (iii)
repeating operation (ii) for remaining coefficients until a test
value corresponding to a coefficient with a lowest order is greater
than or equal to the second threshold; (iv) deactivating a
coefficient with a highest order from the coefficients if a
corresponding test value is smaller than a third threshold; and (v)
repeating operation (iv) for remaining coefficients until a test
value corresponding to a coefficient with a highest order is
greater than or equal to the third threshold.
32. The system of claim 31 wherein each of the test values is equal
to a partial sum of squared magnitudes of the complex correlator
outputs over a portion of the multi-path intensity profile, a
middle test value corresponding to a middle portion of the
multi-path intensity profile, the middle test value being one of
the test values.
33. The system of claim 32 wherein the equalization characteristic
is equal to a ratio of the sum of the test values to the middle
test value.
34. The system of claim 32 wherein the first threshold is a number
greater than 1.
35. The system of claim 32 wherein the second threshold is equal to
a product of the middle test value and a number smaller than 1.
36. The system of claim 32 wherein the third threshold is equal to
a product of the middle test value and a number smaller than 1.
37. The system of claim 20 further comprising a training module
coupled to the equalizer, the training module updating the
coefficients of the equalizer.
38. The system of claim 37 wherein the training module trains the
active coefficients using a least mean squares algorithm.
Description
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates generally to equalizers for digital
receivers, and more particularly, to a transversal equalizer
trained with a least mean squares (LMS) algorithm for use in a
demodulator for digital radio receivers operating in severe
noise/interference environment with multi-path propagation
conditions.
2. Description of Related Art
Propagation conditions in mobile communications are often described
as a multi-path fading channel model. In this model, the signal
arriving at a receiver is a combination of rays, each of the rays
is a replica of the original signal at the output of the
transmitter but has a different strength and a different epoch. The
relative delay between the first ray and the last ray can be
several symbol (or information bit) intervals. For example, the
delay can be up to 5 symbol intervals in Global Systems for Mobile
communications (GSM).
In order to compensate for a largely dispersive multi-path fading
channel, demodulators having a Viterbi equalizer (which utilizes
Maximum Likelihood Sequence Estimation) are widely used (e.g., U.S.
Pat. No. 5,091,918 granted to Wales, U.S. Pat. No. 5,285,480
granted to Chennakeshu et al. and U.S. Pat. No. 5,581,581 granted
to Sato).
The Viterbi equalizer uses a propagation channel estimator to
generate all possible signal sequences which could result from
being transmitted through the estimated propagation channel. Due to
the channel memory length, the number of such signal sequences is
limited. These generated signal sequences are then compared with a
received signal sequence. The generated signal sequence which is
closest in code distance to the received signal is selected. The
data sequence associated with the selected signal sequence is the
recovered data sequence.
Propagation channel estimation is performed by detecting a known
data sequence called midamble which is embedded in the middle of a
signal burst. Good estimation of the propagation channel is crucial
for the performance of the Viterbi equalizer. When the propagation
channel becomes very noisy and/or there are strong interfering
signals (both cases are common in the mobile communication
environment), the performance of the Viterbi equalizer can be
significantly compromised since good estimation of the propagation
channel is not available. Furthermore, the Viterbi equalizer is
complicated to implement.
It is known that a transversal equalizer is much easier to
implement than the Viterbi equalizer, and that a transversal
equalizer can handle a large offset if trained with the LMS
algorithm. However, if the coefficients of the transversal
equalizer are not properly initialized, the equalizer needs to be
trained longer for convergence, i.e., its coefficients will take
longer to converge to proper values. For this reason, the LMS
algorithm is in general considered to be slow in converging.
Accordingly, there is a need for a system for configuring and
initializing a transversal equalizer such that it can perform as
well as or better than a Viterbi equalizer in severe noise and/or
interference environment with multi-path propagation
conditions.
SUMMARY OF THE INVENTION
The present invention is a method and a system for configuring and
initializing an equalizer. The equalizer equalizes a signal
received over a propagation channel. The received signal includes
symbols, a synchronizing signal and is formed by one or more rays
including a dominant ray. The method comprising the following: (a)
generating a multi-path intensity profile of the propagation
channel from the received signal; (b) determining an arrival time
of the dominant ray and a propagation channel response
corresponding to the arrival time based on the multi-path intensity
profile; (c) decimating the received signal synchronously with the
arrival time of the dominant ray to provide signal samples to the
equalizer; (d) configuring the equalizer based on the multi-path
intensity profile to deactivate some of the coefficients of the
equalizer; and (e) initializing active coefficients of the
equalizer based on the propagation channel response.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram of a demodulator of the present
invention.
FIG. 2 illustrates the structure of the modulating bits in a time
slot in GSM.
FIG. 3 is a diagram in more details of a profile generator shown in
FIG. 1.
FIG. 4 is a diagram in more details of the correlator shown in FIG.
3.
FIG. 5 illustrates an exemplary multi-path intensity profile (MIP)
of the propagation channel as a function of the time indices of the
outputs of the correlator.
FIG. 6 is a diagram in more details of a determination module shown
in FIG. 1.
FIG. 7 illustrates a transversal equalizer shown in FIG. 1 at the
start of the training mode.
FIG. 8 is a flowchart of the coefficient-reduction process.
FIG. 9 is a diagram of the I-Q combiner for a GSM system.
DESCRIPTION OF THE INVENTION
In the following description, for purposes of explanation, numerous
details are set forth in order to provide a thorough understanding
of the present invention. However, it will be apparent to one
skilled in the art that these specific details are not required in
order to practice the present invention.
The present invention is a system for configuring and initializing
a transversal equalizer. Using a LMS algorithm for training, a
transversal equalizer can handle a large offset, even when the
coefficients of the equalizer are not properly initialized. In such
a case, the equalizer needs to be trained longer, thus its
coefficients will take longer to converge to proper values. For
this reason, the LMS algorithm is in general considered to be slow
in converging. However, with proper configuration and
initialization of the coefficients and timing reference, as
provided by the system of the present invention, the transversal
equalizer can converge fast enough to achieve a performance close
to or better than that of a Viterbi equalizer in a noisy and/or
strong interference environment such as the one near the specified
sensitivity level and reference interference level described in the
standard ETSI/GSM 05.05 for GSM systems. The transversal equalizer
is also much less complicated than a Viterbi equalizer, thus,
easier to implement.
FIG. 1 shows a block diagram of the demodulator of the present
invention. Referring to FIG. 1, the demodulator 100 is preceded by
a converter 1 and followed by an optional I-Q combiner 3 (I-Q
denotes in phase and quadrature).
The converter 1 includes all conversion modules needed to obtain,
from a received radio frequency (RF) or intermediate frequency (IF)
signal, a pair of baseband In-phase and Quadrature (I-Q) digital
samples, denoted by R(nT.sub.s)=I(nT.sub.s)+j Q(nT.sub.s) in a
complex number form, where T.sub.s =T.sub.b /M is the sampling
duration and T.sub.b is the bit interval. In GSM system, the bit
interval T.sub.b is about 3.69 s.
The demodulator 100 demodulates the baseband I-Q signal and outputs
a complex demodulated signal. The optional I-Q combiner 3 converts
the complex demodulated signal to a one-dimensional sequence. The
I-Q combiner 3 is used when the present invention is applied to a
GSM system. It is noted that the optional I-Q combiner 3 may also
be placed inside the demodulator 100, between the transversal
equalizer 25 and the decision module 28.
GSM uses the Time Division Multiple Access (TDMA) method for
multiple access, and Gaussian Minimum Shift Keying (GMSK) as
modulation process. In GSM, a GMSK modulated signal is transmitted
in burst format in an assigned time slot within a TDMA frame. Since
the GMSK modulation is a binary modulation scheme, the symbol
interval is equal to the bit interval T.sub.b.
FIG. 2 shows the structure of bits in a burst to be transmitted in
a time slot. There are 116 data bits which are separated into two
parts by a synchronizing sequence of 26 bits. The data bits are
preceded by 3 tailing bits and succeeded by 3 tailing bits and a
guarding period of 8.25 bits. The synchronizing sequence is also
called the midamble, due to its location in a burst. This burst is
differentially encoded to produce a burst of modulating bits which
is then inputted into the GMSK modulator. The output of the GMSK
modulator is a carrier signal modulated by the burst of modulating
bits. The demodulator 100 receives and demodulates the modulated
carrier signal to recover the 116 data bits of the burst.
The baseband demodulator 100 comprises a sample buffer 20 to store
a set of baseband samples of a received burst, a profile generator
22, a determination module 24, a configuring module 26, an
initializing module 28, a training module 30, a decimator 32, a
transversal equalizer 34, a power measurement module 36, a
Signal-to-Noise-Ratio (SNR) estimator 38, a decision module 40, and
a timing recovery module 42.
The sample buffer 20 stores the baseband I-Q samples collected at a
rate 1/T.sub.s which is a multiple of the bit rate 1/T.sub.b. The
start time to collect samples of the received burst signal is
controlled by the timing recovery module 42.
The profile generator 22 comprises a correlator 221 and a squared
magnitude module 223, as shown in FIG. 3.
FIG. 4 is a diagram of the correlator 221. The correlator 221 is a
finite impulse response (FIR) filter implemented with a T.sub.s
-spaced tapped delay line having only L non-zero coefficients,
where 1/T.sub.s =M/T.sub.b. The L coefficients are the complex
conjugates of the I-Q samples corresponding to the midamble portion
of the received baseband signal at the sampling instants of
interval T.sub.b. At each cycle of T.sub.s, the correlator 221
correlates L received I-Q samples. These L I-Q samples are spaced
apart by T.sub.b in the sample buffer 12.
Under multi-path propagation conditions, the relative delay between
the first ray and the last ray of the radio signal can be over a
number of symbol durations. Typically, this number in GSM system is
about 5. Assuming that NT.sub.b is the maximum relative delay
between the first ray and the last ray, the midamble search window
of the correlator 221 is of length 2NT.sub.b. This length allows
the alignment of a nominal time with the center of the search
window. Therefore, the time span of the related samples in
correlation is (2N+L)T.sub.b and the total number of the output
values is 2MN. The nominal time which is aligned with the center of
search window is provided by the timing recovery module 42.
The squared magnitude module 223 receives the complex correlator
outputs from the correlator 221, and computes the squared
magnitudes of the complex correlator outputs over the midamble
search window. These squared magnitudes of the complex correlator
outputs form a multi-path intensity profile (MIP) of the
propagation channel as a function of the time indices of the
complex correlator outputs. FIG. 5 illustrates an exemplary MIP of
the propagation channel.
The auto-correlation function of the midamble signal is assumed to
be ideal. If the received signal is formed by only one ray, then
the MIP has only one peak and the time index of the peak
corresponds to the arrival time of the single ray. If there are
several rays with different arrival times, then there are several
peaks, the values of which represent the strengths of the rays and
the time indices of the peaks correspond to arrival times of the
rays. The maximum magnitude value corresponds to the strength of
the dominant ray of propagation path. If there is only one ray, the
output of the midamble correlator which corresponds to the dominant
ray is the propagation channel response. If there are several rays,
then the output of the midamble correlator which corresponds to the
dominant ray is an approximation of the propagation channel
response. For example, if there is only one ray and the propagation
channel response is C, and if the received signal is scaled by
C*/.vertline.C.vertline..sup.2, where C* is the conjugate of C,
then the resulting signal can be perfectly demodulated without any
equalization.
Referring to FIG. 6, the determination module 24 comprises a peak
detector 241 and a time determinator 243. The peak detector 241
determines a maximum value of the squared magnitudes of the complex
correlator outputs. The time determinator 243 determines a time
index which corresponds to the maximum value, and the propagation
channel response. This time index is the arrival time of the
dominant ray and the propagation channel response is the complex
correlator output which corresponds to this time index.
The decimator 32 has a decimating rate of M:1. The epoch of the
decimator 32 is aligned with the arrival time of the dominant ray
as determined by the time determinator 243. Decimation on
R(nT.sub.S) results in a sub-sequence R((mM K)T).sub.S =R((mT.sub.b
KT.sub.S) for m=0,1,2, . . . where K is related to the epoch of the
decimator 32 and is an integer from the set {0, 1, . . ., M-1}.
It is important to note that the nominal time, as determined by the
time recovery module 42, and the epoch of the decimator 32, as
determined by the time determinator 243, are not necessarily the
same. For this reason, timing of each burst at the decimator 32 can
be offset from the nominal timing at the sample buffer 20 by any
fraction of symbol duration in a resolution as small as 1/M symbol
duration, where M is the decimation factor of the decimator 32. For
example, if the sample buffer operates to produce 8 samples per
symbol, and if the decimator 32 outputs one sample per symbol,
i.e., if M is 8, then the timing offset can be 0, 1/8, . . . , 6/8
or 7/8 symbol duration. This novel feature of the present invention
allows accurate determination of the time location of the dominant
ray. This in turn allows proper configuration and initialization of
the transversal equalizer, which in turn facilitates fast
convergence.
FIG. 7 illustrates the transversal equalizer 34 at the start of the
training mode. Referring to FIG. 7, the transversal equalizer 34 is
a T.sub.b -spaced tapped delay line equalizer in training mode with
2N+1 coefficients. The input signal of the equalizer 34 is
r(i)=R(iT.sub.b +KT.sub.S).
The equalizer 34 has initially 2N+1 coefficients, as determined by
the configuring module 26. The configuring module 26 includes a
module to determine an equalization characteristic of the MIP, and
an optimizer to deactivate some of the coefficients when the
equalization characteristic is less than a threshold. Thus, after
being configured by the configuring module 26, the number of active
coefficients of the equalizer 34 may be reduced from the initial
number of 2N+1.
The configuring module 26 determines how many coefficients are
needed to ensure good performance based on the estimated dispersion
of the propagation channel, as provided by the MIP. After the
configuration, some of the initial coefficients are excluded for
the training process, i.e., these coefficiens will not be updated
from their initial null values during the training process. The
following discussion describes the coefficient reduction process
used by the configuring module 26. The process can be performed by
a software module.
Based on the MIP, the 2MN samples of the squared magnitudes are
divided into 2N+1 subsets. An array mip[.] is defined by summing up
M samples of squared magnitudes and assigning the sum to a
corresponding element of the array. It is noted that mip[N] is the
sum of M samples in the middle of the 2MN samples, and that mip[O]
and mip[2N] have only M/2 samples in each summation. Each element
of the mip[.] array corresponds to a coefficient of the transversal
equalizer 34.
If the coefficient reduction process is performed by software, then
the following C program can be used to implement it. In the
following C program, start_tap is defined as the index of the first
non-zero coefficient and end_tap as the last non-zero coefficient:
##EQU1##
where p is a scale factor to establish a threshold for the
coefficient reduction process to operate when the propagation
channel is not too dispersive, and are aggressiveness factors and
are both less than 1. The larger the aggressive factors are, the
more aggressive the coefficient reduction is. The non-causal
coefficients, i.e., coefficients with indices from 0 to N, and
causal coefficients, i.e., coefficients with indices from N+1 to
2N+1, can be treated differently by having, and different.
FIG. 8 is a flowchart of the coefficient reduction process 800.
Referring to FIG. 8, in block 802, the process 800 initializes the
index start_tap of the first non-zero coefficient to be 0, and the
index end_tap of the last non-zero coefficient to be 2N, 0 being
the lowest order and 2N being the highest order. This means that
the initial number of coefficients is 2N+1.
In block 804, the process 800 determines whether the equalization
characteristic ##EQU2##
is less than p, a threshold for the coefficient reduction process
to continue to operate. The threshold p is a number greater than
1.
If the equalization characteristic ##EQU3##
is equal or greater than p, this indicates that the propagation
channel is too dispersive or that there is no dominant ray. This
occurs when the MIP is substantially flat, and there is no clear
peak indicating arrival of a dominant ray. In this case, the
coefficient reduction process 800 terminates (block 806). Thus, the
number of coefficients remains 2N+1, as initially.
If the equalization characteristic is less than p, this indicates
that the propagation channel is not too dispersive. The coefficient
reduction process 800 then operates from both ends of the set of
coefficients.
In block 808, the process 800 determines whether mip[start_tap] is
less than mip[N], being a number less than 1. If it is, then the
process 800 deactivates the current coefficient with index start
tap, sets the value of start tap to be the next higher index (block
810), and repeats operation in block 808 with the next mip array
element, which corresponds to the next higher coefficient. By
deactivating a current coefficient in block 810, the process 800 in
effect reduces the number of active coefficients by 1.
If mip[start_tap] is not less than mip[N], then the process
determines that the current value of start tap is the index of the
first non-zero coefficient (block 812). In other words, the current
coefficient is the first non-zero coefficient, starting from the
lowest ordered coefficient. The process 800 then terminates the
coefficient reduction from this lower order end of the set of
coefficients.
In block 814, operating at the higher order end of the mip array,
corresponding to the higher order end of the set of coefficients,
the process 800 determines whether mip[end_tap] is less than
mip[N], being a number less than 1. If it is, then the process 800
deactivates the current coefficient with index end_tap, sets the
value of end_tap to be the next lower index (block 816), and
repeats operation in block 814 with the next mip array element,
which corresponds to the next lower coefficient. By deactivating a
current coefficient in block 816, the process 800 in effect reduces
the number of active coefficients by 1.
If mip[end_tap] is not less than mip[N], then the process 800
determines that the current value of end.sub.tap is the index of
the last non-zero coefficient (block 818). The process 800 then
terminates the coefficient reduction from this higher order end of
the set of coefficients.
The process 800 can either operate concurrently from both ends of
the set of coefficients, or sequentially from one end to the next.
The aggressiveness factors and used by the process 800 are not
necessarily equal. After the process 800 terminates, the number of
active coefficients of the configured equalizer 34 may be much
smaller than 2N+1. This reduced equalizer length will reduce the
processing time needed by the equalizer 34 in demodulating a data
burst.
The initializing module 28 initializes the remaining active
coefficients of the equalizer 34 by equating all of the active
coefficients to zero, except for the coefficient C.sub.N which is
equated to the propagation channel response, as determined by the
time determinator 243. It is noted that any coefficient can be
selected to be equal to the propagation channel response. C.sub.N
is selected for ease of implementation.
The training module 28 trains the initialized coefficients of the
equalizer 34 using the well-known least-mean-squares (LMS).
Referring to FIG. 7, the updated coefficients are computed
according to the following formula: ##EQU4##
where I.sub.k and I.sub.k.sup. are the k-th symbol of the midamble
and the estimated received k-th symbol of the midamble,
respectively, and is the step size of the updating process.
The index j may assume values of a smaller range as a result of
coefficient reduction. C.sub.N+j.sup.(k) is the coefficient after
the (k-1)-th training and its final value is C.sub.N+j. After
training, all coefficient coefficients remain fixed during
demodulation of the current received burst. Thus, after being
trained by the training module 28, the equalizer 34 is simply a FIR
filter.
The power measurement module 36 calculates the average signal power
of the midamble portion, which is an average of I.sup.2 +Q.sup.2 of
the related I-Q samples outputted from the sample buffer 20. This
average power includes noise power and/or interference power. Thus,
it can be written as S+N, where S is the desired signal power and N
is the power of undesired signals including noise and/or
interference.
The SNR estimator 38 calculates S/(S+N-S), where S is the wanted
signal power. S can be calculated from the output of the correlator
221 since the correlator 221 acts as a matched filter to the
midamble and rejects most of the noise and interference.
The decision module 40 is a converter which scales the output of
the equalizer 34 by the estimated SNR and transforms the scaled
output to samples with a given bit-width.
The optional I-Q combiner 3 is illustrated in FIG. 9. The I-Q
combiner 3 is applicable for the demodulation in GSM systems. It
converts the complex representation of the signal to a
one-dimensional sample sequence. The sign bit of each sample in the
sequence is a data bit by hard decision. For the GSM application,
the I-Q combiner 3 can also be placed between the transversal
equalizer 34 and the decision module 40.
The timing recovery unit 42 collects the arrival times of the
dominant rays of a number of bursts, as determined by the
determination module 24 and calculates an average arrival time of
dominant rays. This average arrival time is designated as the
nominal time. Only the arrival times associated with "good" bursts
are used in the averaging process, where "good" means that the
estimated SNR of a burst, as determined by the SNR estimator 38, is
large enough. The timing recovery unit 42 then compares the new
nominal time with the previous nominal time and adjusts as
necessary the clock/counter of the sample buffer 20 to ensure that
the sample buffer 20 will store the desired samples. The nominal
time thus controls the starting time of the storing of the desired
samples arriving from the convertor 1.
The timing recovery unit 42 also provides the nominal time to the
correlator 221 so that the search window for the midamble can be
aligned properly, and to any other units that need the nominal
time.
While this invention has been described with reference to
illustrative embodiments, this description is not intended to be
construed in a limiting sense. Various modifications of the
illustrative embodiments, as well as other embodiments of the
invention, which are apparent to persons skilled in the art to
which the invention pertains are deemed to lie within the spirit
and scope of the invention.
* * * * *